Creating new file system types

If it is necessary to create a new type of file system, file system helpers and mount helpers must be created.

This section provides information about the implementation specifics and execution syntax of file system helpers and mount helpers.

File system helpers

To enable support of multiple file system types, most file system commands do not contain the code that communicates with individual file systems. Instead, the commands collect parameters, file system names, and other information not specific to one file system type and then pass this information to a back-end program. The back end understands specific information about the relevant file system type and does the detail work of communicating with the file system. Back-end programs used by file system commands are known as file system helpers and mount helpers.

To determine the appropriate file system helper, the front-end command looks for a helper under the /sbin/helpers/vfstype/command file, where vfstype matches the file system type found in the /etc/vfs file and command matches the name of the command being executed. The flags passed to the front-end command are passed to the file system helper.

One required file system helper that needs to be provided, called fstype, does not match a command name. This helper is used to identify if a specified logical volume contains a file system of the vfstype of the helper.

The helper returns 0 if the logical volume does not contain a file system of its type. A return value of 0 indicates the logical volume does not contain a log.
The helper returns 1 if the logical volume does contain a file system of its type and the file system does not need a separate device for a log. A return value of 1 indicates the logical volume does contain a log.
The helper returns 2 if the logical volume does contain a file system of its type and the file system does need a separate device for a log. If the -l flag is specified, the fstype helper should check for a log of its file system type on the specified logical volume.

Obsolete file system helper mechanism

This section describes the obsolete file system helper mechanism that was used on previous versions of AIX®. This mechanism is still available but should not be used anymore.

File system helper operations

The following table lists the possible operations requested of a helper in the /usr/include/fshelp.h file:

Helper operations	Value
#define FSHOP_NULL	0
#define FSHOP_CHECK	1
#define FSHOP_CHGSIZ	2
#define FSHOP_FINDATA	3
#define FSHOP_FREE	4
#define FSHOP_MAKE	5
#define FSHOP_REBUILD	6
#define FSHOP_STATFS	7
#define FSHOP_STAT	8
#define FSHOP_USAGE	9
#define FSHOP_NAMEI	10
#define FSHOP_DEBUG	11

However, the JFS file system supports only the following operations:

Operation Value Corresponding Command #define FSHOP_CHECK 1 fsck #define FSHOP_CHGSIZ 2 chfs #define FSHOP_MAKE 5 mkfs #define FSHOP_STATFS 7 df #define FSHOP_NAMEI 10 ff

File system helper execution syntax

The execution syntax of the file system helper is as follows:

OpName OpKey FilsysFileDescriptor PipeFileDescriptor Modeflags
DebugLevel OpFlags

OpName

Specifies the arg0 parameter when the program invokes the helper. The value of the OpName field appears in a list of processes (see the ps command).

OpKey

Corresponds to the available helper operations. Thus, if the OpKey value is 1, the fsck (file system check) operation is being requested.

FilsysFileDescriptor

Indicates the file descriptor on which the file system has been opened by the program.

PipeFileDescriptor

Indicates the file descriptor of the pipe (see the pipe subroutine) that is open between the original program and the helper program. This channel allows the helper to communicate with the front-end program.

Example: The helper sends an indication of its success or failure through the pipe, which can affect further front-end processing. Also, if the debug level is high enough, the helper can have additional information to send to the original program.

Modeflags

Provides an indication of how the helper is being invoked and can affect the behavior of the helper, especially in regard to error reporting. Mode flags are defined in the /usr/include/fshelp.h file:

     Flags                                 Indicator   
#define FSHMOD_INTERACT_FLAG                  "i"
#define FSHMOD_FORCE_FLAG                     "f"
#define FSHMOD_NONBLOCK_FLAG                  "n"
#define FSHMOD_PERROR_FLAG                    "p"
#define FSHMOD_ERRDUMP_FLAG                   "e"
#define FSHMOD_STANDALONE_FLAG                "s"
#define FSHMOD_IGNDEVTYPE_FLAG                "I"

Example: The FSHMOD_INTERACT flag indicates whether the command is being run interactively (determined by testing the isatty subroutine on the standard input). Not every operation uses all (or any) of these modes.

DebugLevel

Determines the amount of debugging information required: the higher the debugging level, the more detailed the information returned.

OpFlags

Includes the actual device (or devices) on which the operation is to be performed and any other options specified by the front end.

Sample helper invocations

The actual invocation of the file system helper for fsck -fp /user is as follows:

execl("/etc/helpers/v3fshelpers","fshop_check","1","3","5","ifp",
        "0","devices=/dev/lv02,fast,preen,mounted")

In this example:

The executable execd is /etc/helper/v3fshelpers.
The name appearing in a listing of processes (ps command) is fshop_check.
The operation requested is FSHOP_CHECK as represented by the value "1".
The file system is open on file descriptor "3".
The pipe through which the helper can send information to the front-end program is open on file descriptor "5".
The ModeFlags string is "-ifp", indicating interactive mode plus force and perror modes.
The DebugLevel is 0, so no extra debugging output is returned to the fsck command.
The OpFlags string tells the back-end programs which device the operation is to be performed on (/dev/lv02), the options requested (fast and preen), and notes that the device is mounted. For the fsck command, no changes will be made because the fsck command does not work on mounted file systems.

Another example of how file system helpers are invoked uses the mkfs command. To create a JFS file system on the existing logical volume named /dev/lv02, type the following to create a mount point:

mkfs /junk

If you want to create a file system and know only the device you want to mount it to, type:

mkfs /dev/lv02

In either case, the following file system helper is invoked:

execl ("/etc/helpers/v3fshelpers", "fshop_make", "5", "3", "5", "-ip",\
 "0", "name=/junk,label=/junk,dev=/dev/lv02")

The operation requested is now FSHOP_MAKE. The mode is interactive and perror. The OpFlags string includes both the mount point and the device.

Mount helpers

The mount command is a front-end program that uses a helper to communicate with specific file systems. Helper programs for the mount and umount (or unmount) commands are called mount helpers.

Like other file system-specific commands, the mount command collects the parameters and options given at the command line and interprets that information within the context of the file system configuration information found in the /etc/filesystems file. Using the information in the /etc/filesystems file, the command invokes the appropriate mount helper for the type of file system involved. For example, if the user types the following, the mount command checks the /etc/filesystems file for the stanza that describes the /test file system.

mount /test

From the /etc/filesystems file, the mount command determines that the /test file system is a remote NFS mount from the node named host1. The mount command also notes any options associated with the mount.

An example /etc/filesystems file stanza is as follows:

/test:
        dev              = /export
        vfs              = nfs
        nodename         = host1
        options  = ro,fg,hard,intr

The file system type (nfs in our example) determines which mount helper to invoke. The command compares the file system type to the first fields in the /etc/vfs file. The field that matches will have the mount helper as its third field.

Mount helper execution syntax

The following is a sample of the execution syntax of the mount helper:

/etc/helpers/nfsmnthelp M 0 host1 /export /test ro,fg,hard,intr

Both the mount and unmount commands have six parameters. The first four parameters are the same for both commands:

operation: Indicates operation requested of the helper. Values are either M (mount operation), Q (query operation), or U (unmount operation). The query operation is obsolete.
debuglevel: Gives the numeric parameter for the -D flag. Neither the mount nor the unmount commands support the -D flag, so the value is 0.
nodename: Names the node for a remote mount or a null string for local mounts. The mount or unmount commands do not invoke a mount helper if the nodename parameter is null.
object: Names the local or remote device, directory, or file that is being mounted or unmounted. Not all file systems support all combinations. For example, most remote file systems do not support device mounts, while most local file systems do not support anything else.

The remaining parameters for the mount command are as follows:

mount point: Names the local directory or file where the object is to be mounted.
options: Lists any file system-specific options, separated by commas. Information for this parameter comes from the options field of the relevant stanza in the /etc/filesystems file or from the -o Options flag on the command line (mount -o Options). The mount command also recognizes the -r (read-only) flag and converts it to the string ro in this field.

The remaining parameters for the unmount command are as follows:

vfsNumber: Gives the unique number that identifies the mount being undone. The unique number is returned by the vmount call and can be retrieved by calling the mntctl or stat subroutine. For the mount helper, this parameter is used as the first parameter to the uvmount subroutine call that actually does the unmount.
flag: Gives the value of the second parameter to the uvmount subroutine. The value is 1 if the unmount operation is forced using the -f flag (umount -f). Otherwise, the value is 0. Not all file systems support forced unmounts.